Dystroglycan is associated to the disulfide isomerase ERp57.

Dystroglycan (DG) is an extracellular receptor composed of two subunits, α-DG and β-DG, connected through the α-DG C-terminal domain and the β-DG N-terminal domain. We report an alanine scanning of all DG cysteine residues performed on DG-GFP constructs overexpressed in 293-Ebna cells, demonstrating that Cys-669 and Cys-713, both located within the β-DG N-terminal domain, are key residues for the DG precursor cleavage and trafficking, but not for the interaction between the two DG subunits. In addition, we have used immunprecipitation and confocal microscopy showing that ERp57, a member of the disulfide isomerase family involved in glycoprotein folding, is associated and colocalizes immunohistochemically with β-DG in the ER and at the plasma membrane of 293-Ebna cells. The β-DG-ERp57 complex also included α-DG. DG mutants, unable to undergo the precursor cleavage, were still associated to ERp57. β-DG and ERp57 were also co-immunoprecipitated in rat heart and kidney tissues. In vitro, a mutant ERp57, mimicking the reduced form of the wild-type protein, interacts directly with the recombinant N-terminal domain of both α-DG and β-DG with apparent dissociation constant values in the micromolar range. ERp57 is likely to be involved in the DG processing/maturation pathway, but its association to the mature DG complex might also suggest some further functional role that needs to be investigated.

Introduction

Dystroglycan (DG) is a member of the glycoprotein complex associated to dystrophin and it is encoded by a single gene that is translated in a single polypeptide chain of about 895 amino acids. This precursor is cleaved into two subunits: α-DG, a highly glycosylated peripheral membrane protein, and β-DG, a transmembrane protein. The two subunits are connected through the C-terminal domain of α-DG and the N-terminal extracellular domain of β-DG, establishing non-covalent interactions. α-DG is composed of two globular domains, N- and C-terminal, joined by a central elongated mucin-like domain (rich in proline, serine and threonine) extensively O-glycosylated and essential for the interaction with other extracellular matrix proteins, such as laminin, perlecan, agrin and neurexins. β-DG crosses the membrane and inside the cell interacts with dystrophin, Grb-2 and caveolin-3 among others [1,2].

The post-translational events leading to mature DG are still largely unknown, although defects of DG processing are strictly related to severe congenital muscular dystrophies. For example, genetic defects linked to the eight known or putative glycosyltransferases responsible for the O-glycosylation of α-DG are related to several forms of congenital muscular dystrophy also called dystroglycanopathies [3]. Aside from the O-glycosylation sites, α-DG primary sequence also contains some conserved sites for N-linked glycosylation [4,5]. Although inhibition of the N-linked glycosylation with tunicamycin generates an aberrant targeting of α- and β-DG subunits to the plasma membrane, its blockage does not prevent the cleavage of the DG precursor peptide [6]; on the other hand, mutations hitting the two N-linked glycosylation sites, located within the region containing the Gly–Ser cleavage site, inhibit the proteolytic processing of the precursor [7]. The amino acid region surrounding the DG precursor cleavage site revealed sequence similarity to proteolytic cleavage sites for some proteins containing SEA-modules [7]. It has been proposed that the separation between the two DG subunits occurs through an autocatalytic cleavage within the SEA-module, which is stabilized by a disulfide bridge between Cys-669 and Cys-713 [8]; however the functional significance of the DG cleavage remains obscure [9]. Our previous work showed that site-specific mutations hitting the amino acids responsible for the interaction between the α-DG C-terminal domain and the β-DG ectodomain in vitro hamper the proteolytic processing of the precursor in a DG construct tagged with GFP and overexpressed in 293-Ebna cells [10,11]. Similar results have been obtained by other authors mutating the two cysteine residues belonging to the β-DG ectodomain, suggesting the presence of a disulfide bridge [12].

In the present paper, an alanine scanning involving all the DG cysteine residues has been carried out in order to achieve a deeper insight on the redox state of the thiol groups of DG. We also evidenced an interaction between DG and ERp57/PDIA3, a member of the protein disulfide isomerase family, both in vitro and in vivo. ERp57 is able to catalyze disulfide bridge oxidation or isomerisation of non-native disulfide bridges and it is localized mainly in the endoplasmic reticulum, where it specifically binds to glycoproteins addressed to the secretory pathway during their folding [13]. Many extracellular matrix proteins have been identified as ERp57 substrates, such as integrins, laminins and agrin amongst others [14]. Although ERp57 is mainly involved in its role of chaperone for quality control in the endoplasmic reticulum, it has also been found in other cellular compartments where it fulfils a variety of other functions [15]. For example, ERp57 has been found in the cytosol, where it participates to a multiprotein complex responsible for the activation of the transcription factor STAT3 [16], and in the nucleus, where ERp57 is associated to DNA in a complex with the transcription factor STAT3 [17,18]. Recently, a function of rapid response receptor of vitamin D has been attributed to ERp57, suggesting its involvement in signal transduction [19].

Material and methods

Antibodies

The monoclonal antibody anti-calnexin was purchased from BD Transduction Laboratories, the monoclonal antibody specific for the C-terminus of β-DG(43DAG) from Novocastra (Newcastle, UK), and the anti-GFP agarose beads and anti-GFP-HRP from Miltenyi Biotec (Germany). The anti-ERp57 antibody was prepared by immunizing rabbits with the human recombinant protein and by a partial purification on immobilized protein G (Sigma), as previously reported [20].

DNA manipulation

The full-length cDNA encoding for murine DG was used as a template to generate three DNA constructs by PCR, corresponding to the N-terminal region of β-DG, β-DG(654–750), the N-terminal region of α-DG, α-DG(30–315), and the C-terminal region of α-DG, α-DG(485–630). Appropriate primers were used to amplify the DNA sequences of interest. Single point mutations were introduced into the murine DG gene, cloned in pEGFP vector, using QuikChange site-directed mutagenesis kit (Stratagene®); all DNA constructs were verified by automated sequencing.

The full-length DNA constructs carrying the point-mutations were also used as templates to generate by PCR the DNA constructs needed for the expression of β-DG(654–750) mutants in E. coli (see below), using the same primers employed to amplify the wild-type β-DG(654–750) sequence.

The full-length cDNA encoding for murine DG was used as a template to generate by the gene splicing by overlap extension technique (GENE SOEing) two overlapping DNA constructs, which allowed, by a third PCR reaction, to insert the c-myc epitope upon the triplet encoding for K500 [21].

Human ERp57, devoid of the 24 amino acids N-terminal presequence, was cloned into pET21 plasmid and expressed in E. coli BL21 [17]. To generate a functional mutant protein, the second cysteine residue in each thioredoxin-like redox active site of ERp57 (Cys-57 and Cys-406) was replaced with a serine one by site-directed mutagenesis. ERp57-Cys57Ser and ERp57-Cys406Ser mutants were generated by PCR overlap extension using appropriate primers. Introduced mutations were confirmed by automated sequencing. Single mutant sequences were cloned into the pET21 vector and combined in the same sequence using EcoRV, which cuts either between the two redox sites of the ERp57 sequence and once within the pET21 plasmid. pET21-ERp57-Cys57Ser and pET21-ERp57-Cys406Ser were digested with EcoRV and the DNA fragments obtained were purified and joined to generate the ERp57-Cys57Ser–Cys406Ser double mutant (mutated ERp57).

Protein expression, purification and biotinylation

The DNA constructs encoding for DG recombinant domains were cloned into a bacterial vector, which is appropriate to express the protein as a thioredoxin fusion product, also containing an N-terminal 6His tag and a thrombin cleavage site. The recombinant fusion proteins were expressed in E. coli BL21(DE3) Codon Plus RIL strain and purified using nickel affinity chromatography. The domains of interest were obtained upon thrombin cleavage. Tricine/SDS-PAGE was used to check the purity of the recombinant proteins under analysis.

The recombinant wild-type and mutant ERp57 were purified by ammonium sulfate fractionation, followed by chromatography steps using the same procedure previously described for the purification of pig liver protein [17,22].

For solid-phase binding assays, recombinant β-DG(654–750) and α-DG(30–315) were biotinylated in 5 mM sodium phosphate buffer at pH 7.4, with 0.5 mg ml−1 sulfo-N-hydroxylsuccinimido-biotin (S-NHS-biotin, Pierce®). The reaction mixture was incubated for 30 min on ice and in the dark and then dialyzed overnight against PBS buffer (2.5 mM KCl, 2 mM KH2PO4, 2 mM Na2HPO4, 140 mM NaCl, pH 7.4). The optimal dilution for signal detection was determined by dot blot analysis and revealed by enhanced chemioluminescence (Pierce®).

Solid-phase binding assays

To assess the binding properties of recombinant ERp57 and its mutant to biotinylated recombinant β-DG(654–750) and α-DG(30–315), and the binding properties of recombinant α-DG(485–630) to biotinylated wild-type and mutant β-DG(654–750), solid-phase binding assays were performed as follows: ≈0.5 μg of ERp57 (or α-DG(485–630)), its mutant and BSA were immobilized on microtiter plates in coating buffer (50 mM NaHCO3 pH 9.6) overnight at 4 °C. After washing with PBS buffer containing 0.05% (v/v) Tween-20, 1.25 mM CaCl2 and 1 mM MgCl2, the wells were incubated with decreasing concentrations of recombinant biotinylated β-DG(654–750) (or α-DG(30–315)) in PBS containing 0.05% (v/v) Tween-20, 3% (w/v) BSA, 1.25 mM CaCl2 and 1 mM MgCl2 for 3 h at room temperature. After washing, the biotinylated bound protein fractions were detected with alkaline phosphatase Vectastain AB Complex (Vector Laboratories®). 5 mg of p-nitrophenyl phosphate dissolved in 10 ml of 10 mM diethanolamine and 0.5 M MgCl2 were used as a substrate for the reaction with alkaline phosphatase. 100 μl of this solution was added to all wells and absorbance values were recorded at 405 nm. For each ligand concentration, the absorbance value (A) originating from coated BSA, was subtracted from the values obtained with coated proteins under analysis.

Cell culture, transfection and protein extraction

293-Ebna cells were grown in DMEM supplemented with antibiotics and 10% (v/v) fetal calf serum. Cells were transfected with 20 μg of wild-type or mutated pEGFPDG using the calcium-phospate method: briefly, DNA was mixed with 125 mM CaCl2 and 50 mM BES (BES buffered saline) and the DNA–calcium phosphate complex was added to the cells. After 24 h cells were treated with 25 mM N-ethylmaleimide (NEM) to preserve mixed disulfides, and lysed in PBS containing 1% Triton X-100, 20 mM NEM and protease inhibitors and centrifuged at 14,000 rpm for 10 min at 4 °C. Cell lysates were subjected either to immunoprecipitation or WGL-enrichment assay and Western blot analysis.

Immunoprecipitation

Equal amounts of total protein extracts from homogenized tissues or from untransfected cells or overexpressing wild-type or mutated DG-GFP constructs were precleared with protein A/G-Sepharose beads (GE Healthcare) at 4 °C for 1 h. Supernatants were immunoprecipitated for 2 h at 4 °C with a specific antibody or with a non related antibody as negative control. The protein–antibody complexes were pulled down by adding protein A/G-Sepharose beads. Sample pellets were then subjected to several washes in PBS. The immunocomplexes were recovered from the protein A/G-Sepharose beads by boiling the samples in electrophoresis loading buffer. Alternatively, cell lysates were immonoprecipitated with anti-GFP conjugated to magnetic beads (Miltenyi) following manufacture's instructions. One percent of total protein extracts used for immunoprecipitation was loaded as input. The immunocomplexes were analyzed by Western blot.

WGL-enrichment assay

Total protein extracts from untransfected cells or overexpressing wild-type or mutated DG-GFP constructs were incubated with succinylated Wheat Germ Lectin (sWGL) Sepharose 6 MB (Amersham) and equilibrated in PBS containing 1% Triton X-100 overnight at 4 °C. After extensive washing with washing buffer (WB) (50 mM Tris–HCl pH 7.8, 500 mM NaCl, 0.1% Triton X-100), bound glycoproteins were eluted in WB containing 300 mM N-acetylglucosamine and analyzed in Western blot.

Electrophoresis and Western blot analysis

Samples for SDS-PAGE were resuspended in SDS-PAGE sample buffer and 50 mM dithiothreitol (DTT) was added to reduce samples where indicated. All samples were boiled for 5 min before electrophoresis. Proteins were resolved on SDS-PAGE and transferred to nitrocellulose membranes for Western blot analysis. Membranes were extensively blocked with 3% milk in PBS containing 0.5% (v/v) Tween-20 (TPBS). Primary antibody incubations were performed overnight at 4 °C in TPBS containing 3% (w/v) BSA, whereas secondary antibodies (polyclonal goat anti-rabbit, UCS Diagnostic®, Italy (1:10000), or monoclonal goat anti-mouse, GenScript Corp ® (1:5000), immunoglobulins conjugated to horseradish peroxidase), were diluted in TPBS containing 3% (w/v) BSA at room temperature for 1 h. The reactive products were revealed using the luminol-based ECL system (Pierce®, USA). Every experiment has been repeated at least three times.

High-resolution fluorescence microscopy

Transiently transfected 293-Ebna cells were fixed and permeabilized in PBS containing 1% Triton X-100 at room temperature for 20 min followed by incubation with the anti-ERp57 antibody (1:50) in PBS containing 0.1% Triton X-100 and 3% BSA for 1 h. After several washes the cells were incubated with anti-rabbit Alexa Fluor633 secondary antibody (Invitrogen) (1:500), imaged with a confocal laser scanning system (TCS-SP2, Leica Microsystems, GmbH, Wetzlar, Germany) and analyzed with ImageJ programs (http://rsbweb.nih.gov/ij/). Laser excitation at 488 nm of the sample was followed by an excitation at 633 nm to collect emission signals from GFP and Alexa Fluor633, respectively.

Results and discussion

Effect of DG cysteine replacement on the DG precursor cleavage

Previous studies carried out in vitro on the recombinant α-DG N-terminal domain indicated the presence of a disulfide bridge between the cysteines Cys-182 and Cys-264, whose disruption dramatically reduced the structural stability of the domain [23]. More recent studies carried out on eukaryotic cells proposed the presence of a disulfide bridge within the β-DG N-terminal domain between the cysteines Cys-669 and Cys-713, which would also represent a structural determinant for the α-DG/β-DG interface [12].

Six DG-GFP constructs were prepared carrying single point mutations in which every DG cysteine was replaced with alanine and used to transiently transfect 293-Ebna cells in order to analyze the effects of such mutations (Fig. 1A). Lysates from 293-Ebna cells overexpressing the mutated DG-GFP constructs were run through an SDS-PAGE, transferred to nitrocellulose and probed with 43DAG, a monoclonal antibody directed against the C-terminus of β-DG, which detected a main band at 60 kDa, corresponding to β-DG-GFP, and a lower band at about 50 kDa due to some N-terminal degradation. Fig. 1(B) shows that the Cys182Ala, Cys264Ala, Cys642Ala and Cys774Ala mutants underwent a correct DG precursor cleavage, whereas the Cys669Ala and Cys713Ala mutations partially hindered the separation between the two DG subunits, as suggested by the presence of a high molecular weight band of about 160 kDa. In fact, this band has the same electrophoretic mobility displayed by the one occurring with the Ser654Ala mutation, located at the physiological α/β maturation cleavage site Gly653–Ser654, that completely hampers the DG precursor cleavage (Fig. 1B) [7,8,11,24].

Fig. 1

Mutations Cys669Ala and Cys713Ala partially hamper the DG precursor cleavage. (A) Schematic representation of DG complex. Numbers on the top refer to the amino acid positions of cysteine residues. The two putative disulfide bridges between the cysteines Cys-182 and Cys-264 and the cysteines Cys-669 and Cys-713 are also reported. SP, signal peptide, TM, transmembrane region. (B) Lysates from 293-Ebna cells transfected with DG-GFP mutants were resolved on SDS-PAGE, transferred to nitrocellulose and probed with a monoclonal antibody directed against the β-DG C-terminus (43DAG). The DG-GFP mutants carrying the point mutations Cys182Ala, Cys264Ala, Cys642Ala and Cys774Ala displayed a band at about 60 kDa corresponding to the cleaved β-DG-GFP. A lower band, at about 50 kDa, probably originated from further proteolysis of β-DG-GFP. The Cys669Ala and Cys713Ala DG-GFP mutants showed a higher band at 160 kDa (black arrow), which is likely to correspond to the unprocessed DG precursor, pre-DG, displayed by the Ser654Ala mutant. The two other lower bands were probably due to the cleaved β-DG-GFP, devoid of its N-terminus. Asterisk marks unspecific bands.

The band corresponding to the cleaved β-DG-GFP, detected with the Cys669Ala and Cys713Ala mutants, showed a faster electrophoretic mobility than the one displayed by all the other constructs, probably due to some further N-terminal degradation.

Effect of Cys669Ala and Cys713Ala mutations on the α-DG/β-DG interaction in vitro

It is known that the two DG subunits form a noncovalent complex involving the C-terminal domain of α-DG and the N-terminal extracellular domain of β-DG [25]. Mass spectrometry experiments revealed that the two cysteine residues, Cys-669 and Cys-713, do not form a disulfide bridge within the recombinant N-terminal extracellular domain of β-DG, β-DG(654–750), expressed in E. coli (data not shown). Nevertheless, in our previous work [10,11,25], we have demonstrated that β-DG(654–750) still interacts with the C-terminal domain of α-DG. In order to evaluate the involvement of the two single cysteines in the α-DG/β-DG interface, two β-DG(654–750) mutants, carrying the two amino acid substitutions Cys669Ala and Cys713Ala, were prepared and their affinity toward the recombinant α-DG C-terminal domain, α-DG(485–630), was tested by solid-phase binding assays. Fig. 2 shows that the affinity of the two mutated domains, used as biotinylated soluble ligands, toward the recombinant α-DG(485–630), coated onto microtiter plates, was fully comparable with the one displayed by the wild-type β-DG(654–750). The mean values of the apparent dissociation constants KD measured were in the micromolar range (4±0.9 μM for Cys669Ala, 6±1 μM for Cys713Ala and 3±0.8 μM for wild-type). This indicates that the two cysteine residues are dispensable for the correct formation of the α-DG/β-DG interface. In contrast with the conclusions of Watanabe and colleagues [12], it can be argued that the inhibition of the DG precursor cleavage, observed with the Cys669Ala and Cys713Ala mutants in transfected eukaryotic cells, is not due to a dramatic impairment of the α-DG/β-DG interface within the DG precursor, but it is likely due to the specific destabilization of a proposed SEA-module, which encompasses the DG precursor cleavage site (a.a. 653–654) and that would be possibly stabilized by the disulfide bridge between Cys-669 and Cys-713 [7,8].

Fig. 2

Mutations Cys669Ala and Cys713Ala do not hinder the interaction between the α-DG C-terminal domain and the β-DG N-terminal domain in solid-phase binding assays. α-DG(485–630) was immobilized on plate, whereas β-DG(654–750) (full circles) and its mutants, β-DG(654–750) Cys669Ala (open circles) and β-DG(654–750) Cys713Ala (open squares), were used as biotinylated ligands. Each continuous line corresponds to a representative experiment (from a set of at least three experiments with similar results) and was obtained by fitting experimental data to a single class of equivalent binding sites equation.

Effect of DG cysteine replacement on the α-DG glycosylation

α-DG⧸β-DG complex can be enriched from eukaryotic cell lysates with wheat germ lectin conjugated to agarose beads. Succinylated wheat germ lectin (sWGL) specifically binds the N-acetylglucosamine moieties covalently linked to the α-DG core protein. As β-DG is associated to α-DG through non-covalent interactions, it can be retained from the beads. In order to verify if the mutations hitting the cysteine residues perturb the interaction between α-DG and β-DG and if they alter α-DG decoration with N-acetylglucosamine residues, each of the six DG-GFP constructs, mutated in a single cysteine residue, was overexpressed in 293-Ebna cells, enriched with sWGL conjugated to agarose beads and analyzed by Western blot. The data obtained indicate that DG cysteine replacement does not alter the α-DG glycosylation pattern. Fig. 3 shows that β-DG-GFP, but not its degradation product, can be detected in sWGL pulled down proteins with the Cys182Ala, Cys264Ala, Cys642Ala and Cys774Ala mutations, demonstrating that α-DG was processed, correctly decorated with N-acetylglucosamine residues and bound to β-DG-GFP. The DG precursors (pre-DGs) obtained with the Cys669Ala and Cys713Ala mutations were pulled down by sWGL, indicating that they are at least partially decorated with N-acetylglucosamine moieties, in accordance with the results obtained with the Ser654Ala mutant and with other DG constructs, carrying mutations hitting the amino acids responsible for the interaction between the two DG subunits and unable to undergo the precursor cleavage [11]. The cleaved β-DG-GFP generated by the Cys669Ala and Cys713Ala mutants, probably devoid of its N-terminus, lost contact with α-DG and was not retained by the sWGL (Fig. 3).

Fig. 3

Western blot of sWGL enriched wild-type and mutated DG-GFP constructs. Wild-type or mutated DG-GFP constructs, overexpressed in 293-Ebna cells, were enriched with sWGL conjugated to agarose beads. sWGL specifically binds to the N-acetylglucosamine moieties covalently linked to the core protein of α-DG. Beads can retain also β-DG-GFP, as it is associated to α-DG through non-covalent interactions, or directly pre-DG. Total protein extracts and protein fractions eluted from sWGL conjugated beads were resolved on SDS-PAGE, transferred to nitrocellulose and probed with 43DAG. Testing wild-type and all mutated DG-GFP constructs, β-DG-GFP was retained by the beads as well as pre-DGs originating from the Cys669Ala and Cys713Ala mutants, in which the cleavage is partially hampered. The cleaved β-DG-GFP originated by these two mutants could not be retained by the sWGL. Tot, total protein extracts, E, protein fractions eluted from sWGL.

Confocal fluorescence microscopy analysis of cysteine mutated DG-GFP constructs

Confocal fluorescence microscopy analysis of 293-Ebna cells overexpressing mutated DG-GFP constructs evidenced that pre-DGs carrying the Cys669Ala and Cys713Ala single point mutations were trapped within ER/Golgi, whereas the Cys182Ala, Cys642Ala and Cys774Ala mutations did not perturb the trafficking neither the membrane localization of the corresponding β-DG-GFP constructs. The Cys264Ala mutant, although still localized at the plasma membrane, showed a more patchy and diffuse cytosolic signal (see Fig. 1 of supplementary data). Differently from the β-DG disulfide bridge, disruption of the putative α-DG disulfide bridge had no tremendous effects on β-DG trafficking and on its membrane localization. This could be explained considering that the folding of α-DG N-terminal domain is largely autonomous [26].

Identification of ERp57/PDIA3, a member of the protein disulfide isomerase family, as a DG associated protein in eukaryotic cells and in rat heart and kidney tissues

We wondered if the formation of the two putative disulfide bridges, within the α-DG N-terminal domain (between Cys-182 and Cys-264) and the β-DG ectodomain (between Cys-669 and Cys-713), may represent a step assisted by a chaperone in eukaryotic cells. Since ERp57/PDIA3 is one of the most popular ER-associated molecular chaperones with oxido-reductase activity that assists many glycoproteins to achieve their final conformation, we explored the possibility that ERp57 may be involved in DG precursor processing and maturation. To this aim, 293-Ebna cells were transfected with a wild-type DG-GFP construct or with the expression vector containing only the GFP construct as control. Before lysis, intact cells were treated with the membrane permeable alkylating agent N-ethylmaleimide (NEM) to prevent disulfide exchange and to reveal possible mixed disulfides between DG-GFP and ERp57. Cell lysates were immunopurified with an anti-GFP antibody conjugated to magnetic beads. Proteins eluted from the beads were resolved on SDS-PAGE under reducing and non-reducing conditions and transferred to nitrocellulose membranes. Western blot analysis was carried out with an affinity purified rabbit polyclonal antibody directed against human ERp57 (anti-ERp57). As shown in Fig. 4(A), anti-GFP co-immunoprecipitated ERp57, which displayed three bands at about 60 kDa, probably related to protein isoforms or modifications, beside a band with a lower molecular mass most likely an ERp57 degradation product. Anti-GFP failed to precipitate ERp57 from 293-Ebna cells overexpressing GFP alone, demonstrating the specificity of the interaction between DG and ERp57. No bands could be detected at higher molecular mass and corresponding to mixed disulfide products between β-DG-GFP and ERp57, thus indicating a non-covalent association. Similarly, anti-ERp57 co-immunoprecipitated β-DG-GFP, as can be observed by Western blot analysis with an anti-GFP antibody conjugated with horseradish peroxidase and with 43DAG (Fig. 4B). To exclude the possibility that the association between β-DG and ERp57 was the result of DG-GFP overexpression in 293-Ebna cells, untransfected 293-Ebna cell lysates were immunoprecipitated with anti-ERp57 and immunoblotted with 43DAG. The presence of a specific band in the immunoprecipitated sample demonstrated that the interaction between ERp57 and β-DG was not induced by DG-GFP overexpression (Fig. 4B). Endogenous β-DG was co-immunoprecipitated by anti-ERp57 also in rat tissue extracts, such as heart and kidney, where a specific band was detected in Western blot analysis carried out with 43DAG (Fig. 4B).

Fig. 4

ERp57 was co-immunopurified with β-DG-GFP. (A) Lysates from 293-Ebna cells, overexpressing GFP alone or DG-GFP, were subjected to immunopurification with anti-GFP and immunoblotted under reducing (+DTT) and non-reducing (−DTT) conditions with anti-ERp57 and anti-calnexin after stripping of the same nitrocellulose. 20% of the immunopurified fractions were immunoblotted with anti-GFP. Anti-ERp57 detected three bands at about 60 kDa related to three ERp57 isoforms, whereas anti-GFP detected GFP alone at about 20 kDa and β-DG-GFP at about 60 kDa. (B) Lysates from 293-Ebna cells, untransfected or overexpressing DG-GFP, or from rat kidney and heart extracts were subjected to immunopurification with anti-ERp57 and immunoblotted with anti-GFP-HRP, 43DAG and with anti-calnexin after stripping of the same nitrocellulose. (C) Lysates from 293-Ebna cells, transfected with a wild-type DG-GFP construct or with a myc tagged DG-GFP-K500-myc construct, were subjected to immunopurification with anti-myc and immunoblotted with anti-ERp57 and 43DAG. (D) Lysates from 293-Ebna cells, untransfected or overexpressing wild-type DG-GFP, were sWGL enriched and then subjected to immunopurification with anti-ERp57 and immunoblotted with anti-GFP-HRP or 43DAG. Tot, total protein extracts, E, protein fractions eluted from anti-GFP, anti-myc or anti-ERp57, eDG, endogenous dystroglycan, sWGL, succinylated wheat germ lectin enriched protein fraction.

Since the lectin calnexin is often associated to ERp57 to assist the folding of its targets [27], we tested the presence of calnexin in the immunoprecipitates. Nitrocellulose membranes of the previous Western blot experiments were stripped and incubated with an anti-calnexin antibody, which failed to detect calnexin in the immunoprecipitated samples (Fig. 4A and B).

There are not available antibodies directed against α-DG expressed in 293-Ebna cells. In order to verify the presence of α-DG in the β-DG-GFP/ERp57 protein complex, a DG-GFP construct was expressed, in which a 10 amino acids myc-tag was inserted before the Ig-like domain of the α-DG C-terminal portion, at the amino acid position 500 [28]. The DG-GFP-K500-myc construct was immunopurified with an anti-myc antibody conjugated to magnetic beads and immunoblotted with rabbit anti-ERp57 and 43DAG, demonstrating the presence of α-DG in the β-DG-GFP/ERp57 protein complex. Co-immunoprecipitated ERp57 displayed a different mobility compared to protein detected in the whole extract and this could be related to the enrichment of a specific protein isoform or modification, as above observed in Fig. 4(A). As control, ERp57 was not detected in the anti-myc immunopurified fractions from cells overexpressing DG-GFP, demonstrating the specificity of the antibodies utilized (Fig. 4C). The presence of α-DG in the β-DG-GFP/ERp57 complex was confirmed using sWGL, which specifically binds α-DG, to enrich the DG complex, and then anti-ERp57 to co-immunoprecipitate sWGL enriched β-DG-GFP or endogenous β-DG (Fig. 4D).

Effect of DG cysteine replacement on DG association to ERp57

In order to determine whether ERp57 is still associated to the mutated DG-GFP constructs, they were overexpressed in 293-Ebna cells and immunopurified with an anti-GFP antibody conjugated to magnetic beads. The eluted proteins were resolved on SDS-PAGE under reducing and non-reducing conditions and transferred to nitrocellulose membrane to detect ERp57 with anti-ERp57. The bands at about 60 kDa referring to ERp57 could be observed with all the mutants tested, indicating that none of the DG cysteine residues appears to be specifically responsible for the ERp57/DG association (Fig. 5A). Besides ERp57, an additional unidentified band of about 160 kDa was detected with the Cys669Ala and Cys713Ala mutants, either under non-reducing or reducing conditions (Fig. 5A). To rule out the possibility that the 160 kDa component was the result of a mixed disulfide between ERp57 and β-DG-GFP, resistant to a strong reducing treatment, we prepared a DG-GFP construct carrying the Cys669Ala–Cys713Ala double mutation. This was immunopurified with anti-GFP and analyzed by Western blot performed with anti-ERp57, under reducing and non-reducing conditions. Fig. 5(A) shows that the signal at 160 kDa could be detected also with the Cys669Ala–Cys713Ala double mutant. At present we cannot attribute the 160 kDa band to any specific cellular product, but it may be hypothesized that the Cys669Ala and Cys713Ala mutants, which are blocked within the ER/Golgi compartment (see Fig. 1 of supplementary data) form some sort of covalent complexes with the chaperone ERp57. In order to disrupt together the two putative disulfide bridges, Cys182–Cys264 and Cys669–Cys713, located within the α-DG and β-DG N-terminal domains, respectively, a DG-GFP construct, carrying the Cys264Ala–Cys669Ala double mutation, was prepared and overexpressed in 293-Ebna cells, immunopurified with anti-GFP antibody conjugated to magnetic beads and immunoblotted with anti-ERp57. About 20% of the immunopurified protein fraction was immunoblotted with 43DAG. Interestingly, the 60 kDa bands related to ERp57 were lost under both, reducing and non-reducing conditions (Fig. 5B). This may suggest that at least one of the two potential disulfide bridges of DG must be maintained for DG to interact with ERp57. Curiously, the same unidentified band at 160 kDa, previously observed, was also visible in the immunoprecipitate of Cys264Ala-Cys669Ala double mutant, confirming that this occurrence is related to the impairment of DG processing.

Fig. 5

Effect of the DG-GFP cysteine replacement on the DG-GFP/ERp57 interaction. 293-Ebna cells, overexpressing wild-type or mutated DG-GFP constructs, carrying the mutations Cys182Ala, Cys264Ala, Cys642Ala, Cys669Ala, Cys713Ala, Cys669Ala–Cys713Ala and Cys774Ala (A) and Cys264Ala–Cys669Ala (B), were treated with NEM before lysis. β-DG-GFP constructs were immunopurified with anti-GFP conjugated to magnetic beads. Protein fractions eluted from the beads were resolved on SDS-PAGE under reducing (+DTT) or non-reducing (–DTT) conditions, transferred to nitrocellulose, and probed with anti-ERp57 or 43DAG. Tot, total protein extracts, E, protein fractions eluted from anti-GFP conjugated to magnetic beads.

DG and ERp57 co-localize within 293-Ebna cells

Confocal fluorescence microscopy, performed on 293-Ebna cells transiently transfected with wild-type DG-GFP and carrying the Cys264Ala–Cys669Ala double mutation, showed that both, wild-type DG-GFP, which is predominantly localized at the plasma membrane, and the Cys264Ala–Cys669Ala mutant, which is retained within ER/Golgi compartments, co-localized with endogenous ERp57 in discrete patches throughout the cells suggesting a possible association of the two proteins both, within ER compartment and at the plasma membrane (see Fig. 2 of supplementary data).

The recombinant α-DG and β-DG N-terminal domains bind only the reduced form of ERp57

ERp57 is formed by four thioredoxin-like domains with the N- and C-terminal domains containing the catalytic CXXC motifs [29]. The N-terminal cysteine within these motifs forms a mixed disulfide with substrates, and the C-terminal cysteine of the motif subsequently attacks the intermolecular disulfide bond, leading to the release of substrate and to the oxidation of the catalytic motif [30]. An ERp57 double mutant, in which the C-terminal cysteine of both catalytic sites was replaced with a serine, simulates the reduced form of the protein. In order to verify a direct interaction between ERp57 and the DG domains containing the putative disulfide bridges, solid-phase binding assays were carried out coating onto a microtiter plate the recombinant ERp57 and its double mutant and using biotinylated recombinant N-terminal domains of both, α-DG and β-DG, as soluble ligands at increasing concentrations (up to 20 μM). Fig. 6 shows that neither α-DG nor β-DG N-terminal domains displayed significant affinity toward wild-type ERp57, whereas they bound to the double mutant with an apparent dissociation constant in the micromolar range (3±0.9 μM for α-DG(30–315) and 10±0.9 μM for β-DG(654–750)). This suggests the existence of a direct non-covalent interaction between DG and ERp57, which depends on the redox state of ERp57. It should be noted that the apparent dissociation constant values measured are fully comparable with the ones measured for the binding between β-DG(654–750) and the α-DG C-terminal domain, its biological partner [10,11].

Fig. 6

Mutated ERp57 binds the N-terminal domains of α-DG and β-DG in solid-phase binding assay. Recombinant wild-type (open circles) and mutated (full circles) ERp57 were immobilized on plate and tested for the binding towards soluble biotinylated α-DG(30–315) (A) and β-DG(654–750) (B). Each continuous line corresponds to a representative experiment (from a set of at least three experiments producing similar results) and was obtained by fitting experimental data to a single class of equivalent binding sites equation.

Conclusions

In this paper we report for the first time an association between ERp57, a member of the disulfide isomerase family, and DG. In the immunopurified protein fractions containing a chimaeric protein DG-GFP and its mutants, Cys182Ala, Cys264Ala, Cys642Ala and Cys774Ala, which are correctly separated into two subunits, glycosylated and targeted to the plasma membrane, ERp57 was detected at its expected relative molecular mass of about 60 kDa. With the Cys669Ala and Cys713Ala mutants, which are entrapped within the ER/Golgi compartment, ERp57 was detected also as a high molecular mass species probably related to stable complexes between ERp57 and pre-DG. No species with higher molecular mass due to mixed disulfides between ERp57 and DG were detected under non-reducing conditions, suggesting a non-covalent interaction between DG and ERp57. However, there is the possibility that mixed disulfides between DG and ERp57 could not be detected because such species are too transient to be accumulated within the cells and revealed with this technique [31]. ERp57 did not bind to DG when it carries the Cys264Ala–Cys669Ala double mutation, which disrupts both the two putative disulfide bridges, Cys182–Cys264 and Cys669–Cys713. It can be argued that ERp57 may assist DG during its post-translational maturation, catalyzing the formation of the two disulfide bridges.

In addition, it can be hypothesized that ERp57 is associated also to the completely mature DG to modulate its redox state. This is supported by two lines of evidence: the first one refers to the co-localization of ERp57 and wild-type DG within the ER/Golgi compartments but also at the plasma membrane where DG is correctly folded (Fig. 2 of supplementary data). The second one comes from the solid-phase binding experiments reported in Fig. 6 showing that the recombinant N-terminal domain of both, α-DG and β-DG, binds only the recombinant ERp57 mutant mimicking the reduced form of the protein. Recently, it has been proposed that enzyme catalyzed reduction or reorganization of disulfide bonds within surface receptors may induce conformational changes capable of modulating their activity [32,33]. Accordingly, it was hypothesized that cell-surface ERp57 regulates platelet aggregation, driving integrin transition from an inactive state to a ligand binding conformer by catalyzing the remodeling of disulfide bonds within the receptor itself or other substrates [34]. Other cell surface receptors were found to be associated to members of the disulfide isomerase protein family, such as the receptor glycoprotein GP1bα, which is associated to PDI on platelets [35] and the integrin β3 subunit, which is associated to ERp5 [36]. In these cases, the disulfide isomerases do not assist their targets during folding, but modulate the redox state of the completely mature protein receptors at the cell membrane. Further experiments will be needed to test this hypothesis.

Funding

FS is a recipient of a fellowship from the Association Française contre les Myopathies (AFM). The research work leading to these results has also received funding from the Italian Telethon Foundation [GGP06225] to AB and the European Community's Seventh Framework Program [FP7/2007–2013, 227764].